*4.10. Statistical Analysis*

## 4.10.1. Metabolic Data

The effects of diet and E2 treatment were analyzed on longitudinal data using mixed model repeated measures ("lme" in R "nlme" package) [113] using mouse as a random subject. Once the main effects were observed, the separate effects during STND and HFD were measured using repeated measures ANOVA (spss v.24) [114]. Data from metabolic cages, including food and water intake, locomotor activity, and respiration (O<sup>2</sup> consumption and CO<sup>2</sup> production) and its derivative measures, respiratory exchange ratio and resting energy expenditure, were recorded over 72 h and averaged to produce 24-h data, and analyzed using the *t*-test. Fasting blood glucose, plasma hormones and cytokines, and end point measures (including clamp data) were analyzed using the *t*-test. *p* < 0.05 was considered statistically significant.

#### 4.10.2. 16S rRNA Sequence Data

The data were rarefied to the minimum depth of 44,869 prior to the α-diversity and β-diversity analyses [115]. For α-diversity analysis (Chao1 richness and Pielou's evenness indices), linear mixed effects models were fitted to the alpha diversity measures with a random intercept for each mouse ("lme" in R "nlme" package). Wald test was used for assessing the significance. For β-diversity analysis (Bray-Curtis distance), PERMANOVA test (R adonis, 1000 permutations) was used to test whether overall microbiota composition is associated with E2 or diet. For testing the E2 effects, the mouse (not individual sample) was the permutation unit; for testing the diet effects, the mouse was the permutation stratum (i.e., permutation only occurred within the same mice) [116]. The R<sup>2</sup> was given as the effect size.

Differential abundance analysis of treatment (E2 vs Veh) and diet (HFD vs STND) effect was performed on the phylum, class, order, family and genus level. Only taxa with prevalence >10% and maximum proportion >0.2% were tested. Generalized linear mixed effects model (R "glmmPQL" function, over-dispersed Poisson regression, random intercept) was fitted to the aggregated counts accounting for within-mouse correlation [117]. The library size was estimated using GMPR method [118]. The log library size was included as an offset in the regression model. Treatment and diet variables were included as covariates. Potential treatment and diet interaction (GxD) was also investigated by including the interaction term in the regression model. Wald test was used to test the significance of the association. Data were winsorized at 95% quantile (i.e., we replace outlier counts with 95% quantile) to reduce the influence of potential outliers. False discovery rate (FDR) control (BH procedure, R p.adjust function) was used for multiple testing correction and performed on each taxonomic level from phylum down to genus. The taxa with an FDR-adjusted *p* value (or q value) < 0.05 were considered as statistically significant.

#### 4.10.3. Correlation Analysis of Microbiome and Metabolic Data

To identify if any changes in E2-dependent metabolic effects significantly associate with changes in gut microbiota, correlation analyses were performed between the two outcomes. PERMANOVA was used to perform an overall association test based on the Bray-Curtis distance. For metabolic measures where multiple samples within the same mouse were obtained, within-mouse permutations were done. Next, correlation tests were done to identify microbial taxa associated with metabolic measures. To control for the potential confounding effects due to diet and E2 treatment, residuals were taken by fitting regression models (linear mixed effects model) to the microbial taxa abundance (square-root transformed) and metabolic measures adjusting for diet and E2 effects. Spearman correlation tests were then performed on the residuals. To reduce multiple testing burden, correlation analyses were focused on the taxa associated with E2 treatment. The associations with an FDR-adjusted *p* value (or q value) < 0.1 were considered as statistically significant.

#### **5. Conclusions**

The present study provides compelling evidence that estrogens profoundly impact energy and metabolic homeostasis in female mice. Consistent with previous studies, the key metabolic changes, including food intake, energy expenditure, and glucose turnover, were improved by E2 in females fed HFD. Moreover, the present findings reveal that gut microbiota and gut barrier integrity are additional targets of E2-mediated protection against diet-induced metabolic disorders. Furthermore, the role of gut microbiota in metabolic health is supported by the present findings of strong correlations of multiple microbial taxa with specific metabolic measures and physical activity. In future studies, it will be important to perform shotgun metagenomics for the functional study of the gut microbiome and explore the potential beneficial effects of *Akkermansia* and other microbes identified in this study and their causative links with metabolic protection in females provided by E2. In addition, identification and characterization of microbial metabolites that contribute to the beneficial effects of E2 on metabolism will provide important insights for targeting gut microbiota to improve women's metabolic health.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/metabo11080499/s1. Figure S1. E2 alters water intake in female mice during STND (A) and HFD (B). Mice were kept in metabolic cages from days 11–13 after ovariectomy and E2 implant and the average 24-h data were obtained from 72-h data and used for statistical analysis. \* indicates differences between E2 and Veh mice (*n* = 6/group) (*p* < 0.05, *t*-test), Figure S2. Estradiol decreases the respiratory exchange ratio (RER) in female mice during the day. Respiratory exchange ratios of mice on STND (A) or HFD (B) were measured in metabolic cages for 72 h. The average 24-h data were obtained from 72-h data and used for statistical analysis. \* indicates differences between E2 and Veh mice (*n* = 6/group) (*p* < 0.05, *t*-test), Figure S3: Estradiol or high fat diet did not alter levels of the plasma cytokines, IL-6 and TNF-α in female mice. 5 h-fasting blood samples were used to measure IL-6 (A) and TNF-α (B) during STND (on D8) and HFD (on D23); *n* = 6/group, Figure S4. Estradiol does not alter hepatic insulin sensitivity and lipid production in female mice on HFD. Mice E2 (*n* = 6) and Veh (*n* = 5) underwent hyperinsulinemic-euglycemic clamp on days 43–45, a week following jugular vein surgery. Blood glucose (A) Whole-body glycolysis (B) Hepatic glucose production (C) Hepatic insulin action (D) Liver triglycerides (E), Figure S5. Estradiol does not alter zona occludens (ZO-1) immunoreactivity in the colonic epithelium in female mice fed a HFD. Percent area (A) and mean intensity (B) in the three subdivisions of the colon (*n* = 6/group), Figure S6. HFD alters gut microbiota α-diversity in female mice. HFD lowers richness (A) and increases evenness (B), (*n* = 12/group). \* indicates a difference between STND vs. HFD (*p* < 0.05; "lme" in regression).

**Author Contributions:** Each author has made substantial contributions to the work: Conceptualization; formal analysis; writing—original draft; supervision, review and editing: K.D.A., J.K.K. and M.J.T. Methodology, review and editing: K.D.A., S.S., H.L.N. and R.H.F. Formal Analysis: K.D.A., C.C.G., A.E.R.P., and J.C. Project administration and funding acquisition: K.D.A., M.E.G., J.K.K. and M.J.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by NIH 5U24DK076169-13 Subaward # 30835-64 (K.D.A.), NIH 5U2C-DK093000 (J.K.K.), and NIH R01 DK61935 and Wellesley College Jenkins Distinguished Chair in Neuroscience Funds (M.J.T.). Part of this study was performed at the National Mouse Metabolic Phenotyping Center (MMPC) at UMass Medical School.

**Institutional Review Board Statement:** The study was conducted in accordance with National Institutes of Health Animal Care and Use Guidelines, and approved by the Institutional Animal Care and Use Committees of University of Massachusetts Medical School (PROTO202000104, 11/01/20) and Wellesley College (#2101, 02/05/21).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Datasets generated during and/or analyzed during the current study are not publicly available but can be made available upon reasonable request.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**

